Research Summary: Extracting Illicit-Rich Subgraph from Transaction Networks via Feed-Guided Shortest Paths

A research study introduces a motif-based coherent feed selection method for the Shortest Paths Network Search Algorithm (SPNSA) to optimize investigative subgraph extraction in Anti-Money Laundering (AML) transaction graphs. Testing on the IBM-AML benchmark and the Elliptic Bitcoin transaction graph showed that this topology-driven approach yields markedly higher illicit entity ratios than random or degree-based heuristics while keeping the extracted subgraphs compact.

Why it matters

This research addresses a fundamental operational challenge in financial crime compliance: isolating suspicious transaction pathways from massive networks without overwhelming investigative resources. Demonstrating that motif-based feed selection significantly enriches illicit entity concentrations enables compliance functions to accelerate alert triage across both conventional banking and cryptocurrency environments.

Key insights

• The efficacy of the Shortest Paths Network Search Algorithm (SPNSA) in AML network analysis is critically sensitive to the choice of the initial entity feed.
• A novel motif-based coherent feed selection method was developed using graph-only structural properties to guide entity selection.
• Empirical validation on the Elliptic Bitcoin network and the IBM-AML benchmark (HI variants) demonstrates that motif-guided selection delivers significantly higher illicit ratios than random or degree-based approaches.
• The proposed method maintains output compactness, avoiding unnecessary graph expansion while concentrating illicit activity for investigative review.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2610.13280